Aurélien Valade

Numerical simulations are a key tool for deciphering the dynamics of gravitation. Yet, they fail to spatially reproduce the Universe we observe, limiting the comparison between observations and simulations to a statistical level. This is highly problematic for rare, faint, or well-studied nearby objects that are observed in a single environment. The computational cost of recovering this environment in random simulations is prohibitive for most realistic applications. We present ̋amletPM, a method that enables the constraining of initial conditions for cosmological simulations so as to produce evolved numerical universes that can be directly compared to observations of the local Universe. Such simulations are known as constrained simulations. Our method implements the field-level forward modeling of the early-time density field from sparse and noisy measurements of late-time peculiar velocities. The dynamics are integrated with a coarse particle-mesh gravity solver, probing the mildly nonlinear regime. We applied our code to the Cosmicflows-4 compilation of peculiar velocities up to z ). The resulting constrained simulations were analyzed under the framework of the local Universe model as well as an amended version that captures systematic spatial shifts in the scattering of simulated counterparts. (160,h^ -1 ̊m Mpc h^ -1 Mpc ^3 box are presented. These spatially reproduce the observed large-scale structure in the constrained volume. Special attention is given to twelve prominent nearby galaxy clusters, whose simulated counterparts are matched on criteria of mass and separation. We provide a mass estimate constrained by the dynamical environment for each cluster. Field-level forward modeling of the initial conditions can be applied to actual observations so as to produce highly constrained cosmological simulations. In its current form, this method already overtakes in quality the pipeline currently in use in the peculiar-velocity community, although systematic biases still need to be addressed. Furthermore, improving the model can be done easily thanks to the inherent flexibility of the Bayesian approach.

We develop an effective description of the local cosmic environment — namely, for redshift z ≲ 0.1 — to quantify the bias induced by local structure on cosmological observables. Our approach models the metric of the nearby Universe as a superposition of multi-structured Λ-Szekeres patches, calibrated against the HAMLET peculiar velocity and density field reconstructions of Cosmicflows-4. From this framework we compute the fully inhomogeneous and anisotropic quasilocal expansion field predicted by our model, and use it to assess the impact of local structure on estimates of H ₀ . For this purpose we analyse low-redshift Type Ia supernovae from the Pantheon+ catalogue. We find that accounting for the local structure increases the Hubble tension, yielding a shift in the best-fit value of the Hubble constant of order Δ H ₀ ≈ 0.5 km s ^-1 Mpc ^-1 .

Unveil the Local Universe: Les galaxies de l’Univers forment un édifice gigantesque et complexe, appelé structure à grande échelle (LSS). Cependant, la grande majorité de la matière est considérée comme sombre, c’est-à- dire qu’elle n’est pas directement observable par nos télescopes et qu’elle n’est détectable que par son interaction gravitationnelle avec son environnement. La relation entre la distribution des galaxies et celle de la matière n’étant à ce jour pas entièrement bien décrite, la LSS ne peut être réduite aux galaxies qui l’habitent. Dévoiler la distribution de la matière et le champ de vitesse associé dans l’Univers local est une tâche extrêmement difficile. Une approche consiste à combiner, pour chaque galaxie, une mesure du décalage vers le rouge et une estimation de la distance pour obtenir sa vitesse par rapport à son environnement local. La seule source de mouvement à ces échelles étant la gravitation, le champ de vitesse est étroitement lié à la distribution de matière, et les deux peuvent être reconstruits ensemble. Cependant, l’estimation des distances, et donc des vitesses, est difficile à réaliser : les données sont rares, entachées d’erreurs et entachées de biais d’observation. Seule la composante radiale de la vitesse peut être mesurée et la taille de l’erreur croît avec la distance. Des méthodes mathématiques puissantes doivent donc être employées. Notre méthode suit l’approche de l’inférence bayésienne développée au cours de la dernière décennie pour pallier les insuffisances de la méthodologie du filtre de Wiener, dont la modélisation simpliste des données nécessite un traitement quelque peu ad hoc des données préalables. La première étape de l’inférence bayésienne consiste à écrire la probabilité conditionnelle d’un ensemble de paramètres d’un modèle donné, à partir d’un ensemble d’observations. La deuxième étape consiste à créer une série de réalisations de cette loi de probabilité à l’ aide d’une méthode de Monte Carlo, sur laquelle des statistiques sommaires peuvent être calculées. Cependant, ce processus est très coûteux en temps de calcul, et les méthodes développées précédemment n’ont pas pu faire face à la taille croissante des problèmes actuels et futurs. Ce travail répond à cette question grâce à deux innovations majeures. Tout d’abord avec l’utilisation de la méthode de pointe Hamiltonian Monte Carlo pour créer les réalisations du postérieur, et avec l’implémentation d’un code accéléré par le GPU qui réduit le temps de calcul par des ordres de grandeur : la reconstruction HAmiltonian Monte carlo de l’environnement localT (Hamlet). Hamlet est d’abord appliqué à des données fictives compatibles avec le modèle mis en œuvre. Nous démontrons que notre méthode converge correctement avec le nombre de contraintes et l’amplitude des incertitudes. Ensuite, Hamlet est exécuté sur un autre catalogue fictif extrait d’une simulation cosmologique de matière noire uniquement. Parallèlement à Hamlet, nous appliquons au même catalogue l’exemple le plus récent de l’approche double canonique : le pipeline Bias Gaussianization correction / Wiener Filter (BGc/WF). Par rapport au BGc/WF, Hamlet est capable d’extraire plus d’informations des données et de produire des cartes plus contrastées. Cependant, de nouveaux biais apparaissent dans sa reconstruction. Enfin, nous appliquons Hamlet à la dernière version des catalogues de vitesses particulières Cosmicflows : Cosmicflows-4. Nous comparons notre reconstruction de la distribution de matière à une large compilation de relevés de décalage vers le rouge, démontrant une correspondance remarquable entre les deux. Nous analysons les bassins d’attraction du champ de vitesse ainsi que le monopôle et le dipôle, montrant une légère tension avec ΛCDM.

Context. As tracers of the underlying mass distributions, the peculiar velocities of galaxies are valuable probes of the Universe, allowing us to measure the Hubble constant or to map the large-scale structure and its dynamics. The catalogs of peculiar velocities, however, are noisy, scarce, and prone to various interpretation biases. Aims. We measured the radial and bulk flow directly from the largest available sample of peculiar velocities and did not impose a cosmological prior on the velocity field. Furthermore, a minimum assumption on the shape of the radial flow at large distances enabled us to estimate the local Hubble constant. Methods. To address these issues, we analyzed the Cosmicflows-4 catalog (CF4), the most extensive catalog of galaxy peculiar velocities, reaching a redshift z = 0.1. Specifically, we constructed a forward-modeling approach assuming only a flat Universe, which reconstructs the radial and bulk flows of the velocity field directly from measurements of peculiar velocities. Our method was tested on a series of 64 simulated catalogs that mimicked the complex selection function of CF4 in space and in magnitude. Based on our mock data, we propose a simulation-based correction method that we applied to the CF4 data. Results. Our method recovers the radial flow and the direction and magnitude of the bulk flow throughout the covered volume without bias. The incompleteness of the data leads to a systematic amplification of the underlying flow, however, which in turn leads to a systematic overestimation of the amplitude of the bulk flow. By estimating the (cosmic) variance of the density field at large distances from the Lambda Cold Dark Matter (ΛCDM) model, we were able to extract a value of 75.9 ± 1 (stat) km/s/Mpc from the radial inflow. With regard to the bulk flow, a 3 σ tension is found with ΛCDM on the supergalactic X direction and on the magnitude of the bulk flow around 140 Mpc/ h and 240 Mpc/ h . In summary, our work confirms the existing tension on the Hubble constant measured locally and a significant tension in the local bulk flow with ΛCDM predictions. More work is needed to handle all the possible sources of systematic errors inherent to the treating of composite catalogs such as CF4.

Numerical simulations are a key tool to decipher the dynamics of gravitation. Yet, they fail to spatially reproduce the Universe we observe, limiting comparison between observations and simulations to a statistical level. This is highly problematic for rare, faint or well studied nearby objects that are observed in a single environment. The computational cost of recovering this environment in random simulations is prohibitive. We present Hamlet-PM, a method that enables the constraining of initial conditions for cosmological simulations so as to produce evolved numerical universes that can be directly compared to observations of the Local Universe: constrained simulations. Our method implements the field-level forward modeling of the early-time density field from sparse and noisy measurements of late-time peculiar velocities. The dynamics are integrated with a particle-mesh gravity solver, thus probing the mildly non-linear regime. The code is applied to the Cosmicflows-4 compilation of peculiar velocities up to z < 0.05 (160 Mpc/h). The constrained ICs a re-simulated with a high precision N-body code. A series of one hundred dark-matter only cosmological constrained simulations with a resolution of 512^3 particles in a 500^3 [Mpc/h]3 box is presented. Special attention is given to twelve prominent nearby galaxy clusters, whose simulated counterparts are matched on criteria of mass and separation. We provide a mass estimate constrained by the dynamical environment for each cluster. Field-level forward modeling of the initial conditions produces highly constrained cosmological simulations. Currently, this method already overtakes in quality the pipeline in use in the peculiar-velocity community, although systematic biases still need to be addressed. Furthermore, improving the model is easy thanks to the inherent flexibility of the Bayesian approach.

Structure in the Universe is believed to have evolved out of quantum fluctuations seeded by inflation in the early Universe. These fluctuations lead to density perturbations that grow via gravitational instability into large cosmological structures. In the linear regime, the growth of structure is directly coupled to the velocity field since perturbations are amplified by attracting (and accelerating) matter. Surveys of galaxy redshifts and distances allow one to infer the underlying density and velocity fields. Here, assuming the LCDM standard model of cosmology and applying a Hamiltonian Monte-Carlo algorithm to the grouped Cosmicflows-4 (CF4) compilation of 38,000 groups of galaxies, the large scale structure of the Universe is reconstructed out to a redshift corresponding to about 30, 000 km/s. Our method provides a probabilistic assessment of the domains of gravitational potential minima: basins of attraction (BoA). Earlier Cosmicflows catalogs suggested the Milky Way Galaxy was associated with a BoA called Laniakea. Now with the newer CF4 data, there is a slight probabilistic preference for Laniakea to be part of the much larger Shapley BoA. The largest BoA recovered from the CF4 data is associated with the Sloan Great Wall with a volume within the sample of 15.5 10^6(Mpc/h)^3, which is more than twice the size of the second largest Shapley BoA.